Transfer learning between crop types for semantic segmentation of crops versus weeds in precision agriculture

Agricultural robots rely on semantic segmentation for distinguishing between crops and weeds to perform selective treatments and increase yield and crop health while reducing the amount of chemicals used. Deep‐learning approaches have recently achieved both excellent classification performance and real‐time execution. However, these techniques also rely on a large amount of training data, requiring a substantial labeling effort, both of which are scarce in precision agriculture. Additional design efforts are required to achieve commercially viable performance levels under varying environmental conditions and crop growth stages. In this paper, we explore the role of knowledge transfer between deep‐learning‐based classifiers for different crop types, with the goal of reducing the retraining time and labeling efforts required for a new crop. We examine the classification performance on three datasets with different crop types and containing a variety of weeds and compare the performance and retraining efforts required when using data labeled at pixel level with partially labeled data obtained through a less time‐consuming procedure of annotating the segmentation output. We show that transfer learning between different crop types is possible and reduces training times for up to 80%. Furthermore, we show that even when the data used for retraining are imperfectly annotated, the classification performance is within 2% of that of networks trained with laboriously annotated pixel‐precision data.     

爆破振动在台阶边坡上的传播规律与预测分析

基于台阶地形模型爆破振动测试实验,研究爆破振动在台阶地形的传播规律及测点位置对爆破振动速度预测的影响规律。结果表明,爆破振动速度随着水平距离的增加总体呈衰减趋势,恒有垂直方向爆破振动速度远大于水平径向和水平切向,台阶边缘爆破振动速度存在局部放大现象;将台阶不同位置测点分成三类:Ⅰ类(包括全部测点)、Ⅱ类(包括水平地面测点、台阶中部、根部测点)、Ⅲ类(包括水平地面测点、台阶根部测点),分别应用萨道夫斯基公式和高程修正公式进行预测,发现高程修正公式的预测精度明显优于萨道夫斯基公式;Ⅲ类测点预测进度最优,Ⅱ类次之,Ⅰ类最差。建议进行边坡地形爆破振动速度预测时,只选取边坡台阶根部测点的振动速度进行预测分析。     

Dairy farmers with larger herd sizes adopt more precision dairy technologies

An increase in the average herd size on Australian dairy farms has also increased the labor and animal management pressure on farmers, thus potentially encouraging the adoption of precision technologies for enhanced management control. A survey was undertaken in 2015 in Australia to identify the relationship between herd size, current precision technology adoption, and perception of the future of precision technologies. Additionally, differences between farmers and service providers in relation to perception of future precision technology adoption were also investigated. Responses from 199 dairy farmers, and 102 service providers, were collected between May and August 2015 via an anonymous Internet-based questionnaire. Of the 199 dairy farmer responses, 10.4% corresponded to farms that had fewer than 150 cows, 37.7% had 151 to 300 cows, 35.5% had 301 to 500 cows; 6.0% had 501 to 700 cows, and 10.4% had more than 701 cows. The results showed that farmers with more than 500 cows adopted between 2 and 5 times more specific precision technologies, such as automatic cup removers, automatic milk plant wash systems, electronic cow identification systems and herd management software, when compared with smaller farms. Only minor differences were detected in perception of the future of precision technologies between either herd size or farmers and service providers. In particular, service providers expected a higher adoption of automatic milking and walk over weighing systems than farmers. Currently, the adoption of precision technology has mostly been of the type that reduces labor needs; however, respondents indicated that by 2025 adoption of data capturing technology for monitoring farm system parameters would be increased.     

Computed microtomography evaluation of calcium hydroxide‐based root canal dressing removal from oval root canals by different methods of irrigation

The aim of this study was to evaluate the capacity of conventional irrigation, passive ultrasonic irrigation (PUI), and Easy Clean for removing calcium hydroxide‐based root canal dressing from oval root canals. Thirty mandibular uniradicular incisors with oval canals were used, and subjected to chemical–mechanical preparation with Reciproc R40 instruments. The main canal was filled with a paste based on Ca(OH)₂ P.A., iodoform and propylene glycol in the ratio of 3:1:1. The teeth were stored in 100% humidity at a temperature of 37°C for 14 days. Afterwards, the teeth were divided into three groups (n = 10) according to the method of irrigation used (conventional irrigation, PUI, and Easy Clean). The specimens were analyzed by computed microtomography at three time intervals: before placing the root canal dressing, with the root canal dressing in place, and after application of the irrigation methods for removing it. The data were submitted to Kruskal–Wallis and Dunn tests for analyzing the canal as a whole, and Friedman and Dunn for analyzing the root thirds. The results showed that conventional irrigation was less efficient for removing the root canal dressing in comparison with the methods that agitated the irrigant solution (p < .05). When the root canal was analyzed as a whole, Easy Clean, and PUI were similar (p > .05). In analysis of the thirds, Easy Clean was more efficient than conventional irrigation in all the thirds, while PUI showed this behavior only in the cervical third (p < .05). The authors concluded that in oval canals, none of the irrigation methods were capable of removing all the root canal dressing, however, the methods that agitated the irrigant solution were more efficient than conventional irrigation.     

Moving beyond ‘more crop per drop’: insights from two decades of research on agricultural water productivity

ABSTRACT

Concern over increasing water scarcity has led to the introduction of the concept of agricultural water productivity and an emphasis on interventions to achieve ‘more crop per drop’. Yet, a strong debate continues on how the concept is to be defined and used. Drawing largely from the irrigation literature, the origins of the concept and its methodological developments are reviewed, and its use in applied work over two decades is discussed. Based on this analysis of conceptual and applied research, key insights into the concept’s contributions and limitations are presented, as well as opportunities for further refinements.     

高频谐振疲劳机载荷测量误差建模分析及试验夹具优化设计

电磁谐振式疲劳裂纹扩展试验过程中，对动态载荷的高精度测量和控制是保证试验结果准确性和系统工作稳定性的重要前提，夹具作为夹持试件和传递载荷的重要机构，对其进行结构优化显得尤为重要。建立了紧凑拉伸(CT)试件连接刚度模型和三自由度有阻尼振动系统动力学模型，得到了动态载荷测量误差理论表达式；分析了夹具刚度对动态载荷测量误差的影响，设计了不同结构刚度的试件夹具，并通过有限元方法计算其刚度，得到了不同结构夹具对动态载荷测量误差的影响结果；实验验证了优化夹具设计的合理性。     

多工作模式压电直线电机的设计与研究

提出了拥有多种工作模式的新型非共振式压电直线电机。对电机的核心部件定子结构进行了设计与建模,分析了其在3种工作模式(连续作动模式、交替步进模式以及单步作动模式)下的作动机理。对电机的布置形式、夹持机构以及预压力机构进行了设计与研究,利用有限元软件辅助分析了结构的机械强度。制作样机并分别在3种工作模式下进行了一系列实验。实验结果证明了多工作模式的可行性:在连续作动模式下施加电压为100V、频率为100Hz的正弦信号,电机输出速度为446.4μm/s;在交替步进作动模式下施加电压为100V、频率为100Hz的方波-三角波信号,电机输出速度为6 031μm/s;在单步作动模式下,电机作动单步步距小于同等条件下交替步进模式步距,当施加电压为30V、频率为1Hz脉冲信号时,其平均步距约为333.33nm。新型直线电机可以适应多种不同场合的工作需求。     

一种高精度大行程2R1T并联定位平台

为了解决传统压电陶瓷驱动器所固有的高精度与大行程的矛盾,提出了一种新型可直接外力驱动的三自由度并联精密定位平台,利用压电直线电机直接驱动即可实现大行程与高精度定位。首先,分析了柔性铰链结构参数对其作动行程的影响机理,在此基础上设计了大行程圆柱柔性铰链,并对大行程圆柱柔性铰链的结构参数进行了模糊优化设计;接着,采用两个平移副和一个一体化、大行程圆柱形柔性铰链构建了并联分支链路,采用3条并联支路设计了2R1T并联精密定位平台,并通过坐标齐次变换方法构建了并联平台的运动学模型,推导出了并联平台位姿正反解;最后,对2R1T并联定位平台开展了实验研究。实验结果表明所设计的2R1T并联定位平台具有良好的运动同步性和重复性,利用压电直线电机的步进运动模式可实现精密定位,其平动定位分辨率为0. 09μm,转动定位分辨率分别为0. 8μrad、0. 9μrad和1. 0μrad;利用压电直线电机的连续运动模式可实现大行程空间定位,其平动工作行程为120 mm,转动工作行程分别为6. 18°、6. 74°和6. 58°。借助于压电直线电机的不同工作模式实现了并联平台的精密定位和大行程工作空间两个关键指标,为进一步研究2R1T并联机构的动态性能、精密定位及控制规律的实现等提供了重要的理论价值和实践基础。